AI Agent Skills

Check AI-written text: extract every verifiable claim, look for sources you can open yourself, and flag likely hallucinations in a structured report.

Guidance for building accessible, responsive interfaces with shadcn/ui and Tailwind CSS, including dark mode, theming and a canvas-based visual design approach.

After a substantive answer you may act on, adds 2-3 specific follow-up questions that help you check facts, probe the reasoning and notice missing context.

Plan A/B tests that give valid results: hypothesis, sample size and duration, metrics, variants and analysis, plus a growth experimentation program with ICE scoring and a playbook.

Convert PDF, Word, PowerPoint, Excel, HTML and other files to clean Markdown with Microsoft MarkItDown, for search, analysis and feeding documents to an AI, with safe local defaults.

Create Playwright automation for inspecting, testing, and debugging local web applications.

Find the real cause of layout bugs at specific screen widths: sideways scrolling, overflow, flex items that will not shrink, sticky that does not stick, 100vh on phones, and fixed bars under the browser UI.

Analyses your meeting transcripts for communication patterns such as conflict avoidance, hedging and speaking ratio, with timestamped examples and better phrasings.

Threat model a code repository: trust boundaries, assets, attacker capabilities, concrete abuse paths and mitigations tied to real files, written up as one concise Markdown report.

Audit a page or component against WCAG 2.2 level AA: keyboard, focus, contrast, names and labels, structure, forms, dialogs, dynamic updates, zoom and target size, with each finding tied to a success criterion.

Write a PRD that links business goals to technical execution: problem, users, measurable requirements, non-goals, specs, risks and a phased roadmap.

Plan, audit, and troubleshoot GA4, GTM, UTM, conversion, and product-event tracking.

Make the agent design landing pages, portfolios and redesigns that do not look templated: read the brief, state a design read, set three dials, pick a real design system when one fits, and pass a strict pre-flight check.

A writing partner for articles and newsletters: outlines, sharper hooks, research with citations and section-by-section feedback.

Design web, mobile and desktop interfaces inside Penpot through its MCP server, with design-system checks, component and accessibility rules and platform sizes.

Fix under-reported ad conversions: capture click ids, carry them to the order, send purchases server-to-server to Facebook, TikTok, Google and Bing, dedupe and verify.

Synthesize interviews, surveys, support data, reviews, and public discussions into evidence-based customer insight.

Turn the current folder into a teaching workspace and learn a topic over several sessions: a stated mission, trusted resources, short interactive HTML lessons, reference sheets and learning records that decide what to teach next.

Find out why an AI image prompt gives messy, wrong or inconsistent pictures and fix it with the fewest changes: contradictions, clutter, mixed styles, ignored details, bad hands and garbled text.

Turn store numbers into a short weekly or monthly report: which metrics to track and how they are defined, why revenue or conversion moved, what is signal and what is noise, and what to do next.

List every state a screen or component needs: loading, empty, error, partial, disabled, offline, permission denied and extreme content, and decide what the user sees and can do in each.

Plan what content to create and why: pillars and topic clusters, keyword research by buyer stage, idea sources, a scoring method for priorities and a 60/30/10 calendar.

A reference for writing documents agents read, such as skills, CLAUDE.md and AGENTS.md, so they behave predictably.

Turn a plain-language description into an Excalidraw diagram file: flowcharts, mind maps, architecture, sequence, ER, class and swimlane diagrams, with templates and helper scripts.